Skip to content

Commit d069fe8

Browse files
keskukesku
committed
showcase folder with vonar example
1 parent 1d8f92d commit d069fe8

File tree

2 files changed

+112
-0
lines changed

2 files changed

+112
-0
lines changed

‎docs/showcase/index.mdx

Lines changed: 92 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,92 @@
1+
---
2+
title: "Community Showcase"
3+
description: "Discover amazing projects built by the community using Perplexity's Sonar API"
4+
sidebar_position: 1
5+
---
6+
7+
# Community Showcase
8+
9+
Discover incredible projects built by developers around the world using **Perplexity's Sonar API**. From voice assistants to research tools, see how the community is leveraging real-time, grounded AI to solve real problems.
10+
11+
## Featured Projects
12+
13+
<CardGroup cols={2}>
14+
<Card title="Voice AI Assistant" icon="microphone" href="/cookbook/showcase/vonar-ai">
15+
Real-time voice assistant for customer support with intelligent call routing and policy queries.
16+
</Card>
17+
18+
<Card title="Submit Your Project" icon="plus" href="https://github.com/ppl-ai/api-cookbook/issues/new?template=showcase-submission.md">
19+
Built something amazing with Sonar? Share it with the community!
20+
</Card>
21+
</CardGroup>
22+
23+
## What Makes a Great Showcase Project?
24+
25+
We love featuring projects that demonstrate:
26+
27+
<Check>
28+
**Real-world impact** - Solves actual problems for users
29+
</Check>
30+
31+
<Check>
32+
**Creative implementation** - Innovative use of Sonar's capabilities
33+
</Check>
34+
35+
<Check>
36+
**Community value** - Inspires or educates other developers
37+
</Check>
38+
39+
<Check>
40+
**Open source** - Code available for others to learn from
41+
</Check>
42+
43+
## Project Categories
44+
45+
### 🎤 **Voice & Conversational AI**
46+
Projects that use Sonar for voice interactions, chatbots, and conversational experiences.
47+
48+
### 🔍 **Research & Analysis**
49+
Tools for academic research, data analysis, and knowledge discovery.
50+
51+
### 🏢 **Business Applications**
52+
Customer support, internal tools, and enterprise solutions.
53+
54+
### 🎨 **Creative & Entertainment**
55+
Games, creative writing tools, and entertainment applications.
56+
57+
### 🛠️ **Developer Tools**
58+
IDE extensions, development utilities, and productivity tools.
59+
60+
## Submit Your Project
61+
62+
Have you built something amazing with Sonar API? We'd love to feature it in our showcase!
63+
64+
<Steps>
65+
<Step title="Prepare your submission">
66+
Ensure your project is publicly accessible and well-documented.
67+
</Step>
68+
69+
<Step title="Create a showcase entry">
70+
Fork the cookbook repository and add your project following our template.
71+
</Step>
72+
73+
<Step title="Submit a pull request">
74+
Open a PR with your showcase entry and we'll review it for inclusion.
75+
</Step>
76+
</Steps>
77+
78+
### Submission Requirements
79+
80+
- **Open source** with accessible repository
81+
- **Working demo** or deployed application
82+
- **Clear documentation** explaining what it does
83+
- **Sonar API integration** prominently featured
84+
- **Professional presentation** with screenshots/demos
85+
86+
## Get Inspired
87+
88+
Browse through the showcase to see how other developers are using Sonar API, learn from their implementations, and get inspired for your next project!
89+
90+
---
91+
92+
**Ready to build?** Check out our [Examples](/cookbook/examples) and [Integration Guides](/cookbook/articles) to get started with Sonar API.

‎docs/showcase/vonar-ai.mdx

Lines changed: 20 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,20 @@
1+
---
2+
title: "Vonar.ai - Voice AI Assistant"
3+
description: "Voice-based AI assistant for customer support using Sonar API for real-time intelligence"
4+
sidebar_position: 2
5+
---
6+
7+
# Vonar.ai - Voice AI Assistant
8+
9+
Vonar.ai is an AI-powered voice agent that answers inbound calls, handles support requests, and instantly finds answers using real-time search — all without a human.
10+
11+
12+
🚀 Inspiration We were frustrated by how bad most inbound call experiences are — long wait times, robotic phone trees, and irrelevant answers. At the same time, most businesses don’t have the resources to build a full AI voice stack. Vonar.ai was born from that gap: we wanted to create a voice agent that actually understands callers, acts fast, and gets smarter with real-time knowledge — without needing a team of engineers to set it up.
13+
14+
⸻ 🤖 What it does
15+
16+
Vonar.ai is a voice-based AI assistant that answers inbound calls, triages requests, and responds using real-time, search-powered intelligence. It can: • Answer questions about policies, outages, or product info using Perplexity’s Sonar API • Handle account, tech, or service-related queries in a guided, secure way • Mock advanced flows like authentication and ticket creation to simulate integrations • Escalate to human agents or inform users when they need a custom production setup
17+
18+
All while speaking naturally and keeping the experience smooth and on-brand. ⸻ 🛠 How we built it • Voice pipeline: Built using Vapi.ai to power voice input/output with real-time LLM processing • LLM: We used Llama 4 via Groq for blazing-fast inference and plugged in Perplexity’s Sonar Pro as a real-time search tool • Tool integration: Designed custom tool calls that silently inject Sonar results into the LLM’s context mid-call • Frontend/Infra: Serverless backend with Netlify Functions, live audio/webhook streaming, and optional Redis caching for performance • Demo optimization: Built a system prompt + dynamic persona system for customized flows per org or use case ⸻ ⚔️ Challenges we ran into • LLM prompt management: Keeping the system prompt concise while packing in real logic and disclaimers for mock/demo flows • Tool call timing: Ensuring that tool responses came back fast enough to avoid Vapi auto-ending the call due to silence • Auth & ticketing mocks: Simulating secure flows in a realistic way without violating trust or misleading the user • Silent lookups: Preventing the AI from “breaking character” when fetching from Sonar — had to tune the tone and behavior closely ⸻ 🏆 Accomplishments that we’re proud of • Built a fully working inbound AI voice agent that can actually hold useful conversations, not just reroute calls • Seamlessly connected real-time search (Sonar) into a voice workflow — almost no one does this live and cleanly • Delivered a usable demo that businesses could test with their own call flow in minutes • Optimized latency with Groq + Netlify + Redis to make everything feel snappy as hell
19+
20+
⸻ 📚 What we learned • Fast voice UX needs ruthless simplicity and aggressive caching — even 500ms of lag ruins the experience • People will forgive a lot if the voice is confident, clear, and useful — tone matters more than people think • Vapi’s infrastructure is powerful but fragile if you miss even one config (like toolIds) • Real-time search + voice agents is a killer combo, especially for gov and compliance-heavy orgs ⸻ 🚧 What’s next for Vonar.ai • Pilot with gov-aligned orgs and service-heavy businesses like MSPs, SaaS platforms, or utilities • Production integrations with Jira Service Management and Salesforce for live ticketing + user auth • Deploy self-serve onboarding so any org can spin up a branded Vonar voice agent with their own FAQ link • Expand tool library to include weather, outage detection, gov databases, or internal knowledge base sync • Go-to-market: targeting Carahsoft channels, AI hackathons, and defense tech conferences for early traction

0 commit comments

Comments
 (0)